Target Person Direction using Motions on Video Sequences of Image-based Avatars

Abstract

We investigate whether an image-based avatar with motions smoothly directs a target person whose order of guide has come. Existing methods using an avatar assumed one-versus-one interaction and did not fully consider how to direct a target person, which is important for interaction with multiple users. The target person feels that the image-based avatar faces to the person and feels that the avatar directs the person, so that the person judges the order of guide has come. The nontarget person feels that the image-based avatar does not face to the person and feels that the avatar does not direct the person, so that the person judges the order of guide has not come. To smoothly direct the target person, the motion of the image-based avatar and the motion of the video sequence are important. We conducted subjective assessment whether the target person and the nontarget person feel about the direction by comparing the cases of adding motions and not adding motions. The results show that the image-based avatar with motions is effective for the target person but is not effective for the nontarget person.